runtime: implement high resolution timer on windows arm/arm64

This CL moves the usleep2HighRes from assembly to good old Go.
This is safe because since CL 288793 usleep is always called with
a g, else one wold have to call usleep_no_g. This condition was
not enforced when high resolution timers were first implemented
on Windows (CL 248699), so the implementation was done in assembly.

Other than removing a bunch of obscure assembly code, this CL makes
high resolution timers work on windows arm/arm64 by free, as the
system calls are the same in all windows platforms.

Change-Id: I41ecf78026fd7e11e85258a411ae074a77e8c7fc
Reviewed-on: https://go-review.googlesource.com/c/go/+/471142
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Michael Pratt <mpratt@google.com>
Reviewed-by: Bryan Mills <bcmills@google.com>
Auto-Submit: Michael Pratt <mpratt@google.com>
Run-TryBot: Quim Muntal <quimmuntal@gmail.com>

src/runtime/os_windows.go

@@ -440,13 +440,7 @@ func createHighResTimer() uintptr {
 		_SYNCHRONIZE|_TIMER_QUERY_STATE|_TIMER_MODIFY_STATE)
 }
 
-const highResTimerSupported = GOARCH == "386" || GOARCH == "amd64"
-
 func initHighResTimer() {
-	if !highResTimerSupported {
-		// TODO: Not yet implemented.
-		return
-	}
 	h := createHighResTimer()
 	if h != 0 {
 		haveHighResTimer = true
@@ -1130,7 +1124,6 @@ func stdcall7(fn stdFunction, a0, a1, a2, a3, a4, a5, a6 uintptr) uintptr {
 
 // These must run on the system stack only.
 func usleep2(dt int32)
-func usleep2HighRes(dt int32)
 func switchtothread()
 
 //go:nosplit
@@ -1152,13 +1145,15 @@ func usleep_no_g(us uint32) {
 //go:nosplit
 func usleep(us uint32) {
 	systemstack(func() {
-		dt := -10 * int32(us) // relative sleep (negative), 100ns units
+		dt := -10 * int64(us) // relative sleep (negative), 100ns units
 		// If the high-res timer is available and its handle has been allocated for this m, use it.
 		// Otherwise fall back to the low-res one, which doesn't need a handle.
 		if haveHighResTimer && getg().m.highResTimer != 0 {
-			usleep2HighRes(dt)
+			h := getg().m.highResTimer
+			stdcall6(_SetWaitableTimer, h, uintptr(unsafe.Pointer(&dt)), 0, 0, 0, 0)
+			stdcall3(_NtWaitForSingleObject, h, 0, 0)
 		} else {
-			usleep2(dt)
+			usleep2(int32(dt))
 		}
 	})
 }

src/runtime/sys_windows_386.s

@@ -241,43 +241,6 @@ TEXT runtime·usleep2(SB),NOSPLIT,$20-4
 	MOVL	BP, SP
 	RET
 
-// Runs on OS stack.
-// duration (in -100ns units) is in dt+0(FP).
-// g is valid.
-TEXT runtime·usleep2HighRes(SB),NOSPLIT,$36-4
-	MOVL	dt+0(FP), BX
-	MOVL	$-1, hi-4(SP)
-	MOVL	BX, lo-8(SP)
-
-	get_tls(CX)
-	MOVL	g(CX), CX
-	MOVL	g_m(CX), CX
-	MOVL	(m_mOS+mOS_highResTimer)(CX), CX
-	MOVL	CX, saved_timer-12(SP)
-
-	MOVL	$0, fResume-16(SP)
-	MOVL	$0, lpArgToCompletionRoutine-20(SP)
-	MOVL	$0, pfnCompletionRoutine-24(SP)
-	MOVL	$0, lPeriod-28(SP)
-	LEAL	lo-8(SP), BX
-	MOVL	BX, lpDueTime-32(SP)
-	MOVL	CX, hTimer-36(SP)
-	MOVL	SP, BP
-	MOVL	runtime·_SetWaitableTimer(SB), AX
-	CALL	AX
-	MOVL	BP, SP
-
-	MOVL	$0, ptime-28(SP)
-	MOVL	$0, alertable-32(SP)
-	MOVL	saved_timer-12(SP), CX
-	MOVL	CX, handle-36(SP)
-	MOVL	SP, BP
-	MOVL	runtime·_NtWaitForSingleObject(SB), AX
-	CALL	AX
-	MOVL	BP, SP
-
-	RET
-
 // Runs on OS stack.
 TEXT runtime·switchtothread(SB),NOSPLIT,$0
 	MOVL	SP, BP

src/runtime/sys_windows_amd64.s

@@ -251,39 +251,6 @@ TEXT runtime·usleep2(SB),NOSPLIT,$48-4
 	MOVQ	40(SP), SP
 	RET
 
-// Runs on OS stack. duration (in -100ns units) is in dt+0(FP).
-// g is valid.
-TEXT runtime·usleep2HighRes(SB),NOSPLIT,$72-4
-	MOVLQSX	dt+0(FP), BX
-	get_tls(CX)
-
-	MOVQ	SP, AX
-	ANDQ	$~15, SP	// alignment as per Windows requirement
-	MOVQ	AX, 64(SP)
-
-	MOVQ	g(CX), CX
-	MOVQ	g_m(CX), CX
-	MOVQ	(m_mOS+mOS_highResTimer)(CX), CX	// hTimer
-	MOVQ	CX, 48(SP)				// save hTimer for later
-	LEAQ	56(SP), DX				// lpDueTime
-	MOVQ	BX, (DX)
-	MOVQ	$0, R8					// lPeriod
-	MOVQ	$0, R9					// pfnCompletionRoutine
-	MOVQ	$0, AX
-	MOVQ	AX, 32(SP)				// lpArgToCompletionRoutine
-	MOVQ	AX, 40(SP)				// fResume
-	MOVQ	runtime·_SetWaitableTimer(SB), AX
-	CALL	AX
-
-	MOVQ	48(SP), CX				// handle
-	MOVQ	$0, DX					// alertable
-	MOVQ	$0, R8					// ptime
-	MOVQ	runtime·_NtWaitForSingleObject(SB), AX
-	CALL	AX
-
-	MOVQ	64(SP), SP
-	RET
-
 // Runs on OS stack.
 TEXT runtime·switchtothread(SB),NOSPLIT,$0
 	MOVQ	SP, AX

src/runtime/sys_windows_arm.s

@@ -211,13 +211,6 @@ TEXT runtime·usleep2(SB),NOSPLIT|NOFRAME,$0-4
 	MOVW	R4, R13			// Restore SP
 	MOVM.IA.W (R13), [R4, R15]	// pop {R4, pc}
 
-// Runs on OS stack.
-// duration (in -100ns units) is in dt+0(FP).
-// g is valid.
-// TODO: needs to be implemented properly.
-TEXT runtime·usleep2HighRes(SB),NOSPLIT|NOFRAME,$0-4
-	B	runtime·abort(SB)
-
 // Runs on OS stack.
 TEXT runtime·switchtothread(SB),NOSPLIT|NOFRAME,$0
 	MOVM.DB.W [R4, R14], (R13)  	// push {R4, lr}

src/runtime/sys_windows_arm64.s

@@ -239,13 +239,6 @@ TEXT runtime·usleep2(SB),NOSPLIT,$32-4
 	ADD	$16, RSP
 	RET
 
-// Runs on OS stack.
-// duration (in -100ns units) is in dt+0(FP).
-// g is valid.
-// TODO: needs to be implemented properly.
-TEXT runtime·usleep2HighRes(SB),NOSPLIT|NOFRAME,$0-4
-	B	runtime·abort(SB)
-
 // Runs on OS stack.
 TEXT runtime·switchtothread(SB),NOSPLIT,$16-0
 	MOVD	runtime·_SwitchToThread(SB), R0